Working in the Environmental & Sustainability area for a cruise line means living at the intersection of ambitious environmental targets and an increasingly complex regulatory landscape. Over the past year, I’ve built two AI-powered tools that tackle very different problems – one aimed at a long-standing operational challenge, food waste, and the other at a reasonably new regulatory requirement, FuelEU Maritime. Neither required a background in AI tools or software development.

Food for thought
Reducing food waste is one of those sustainability goals that sounds simple until you try to act on it. You can’t manage what you can’t measure, and on a cruise ship, “measuring” food waste properly means tracking it across multiple restaurants, several meal services a day, different service styles, and ships that don’t all operate the same way. A buffet breakfast on one vessel behaves nothing like an à la carte dinner service on another, and lumping the data together tells you almost nothing useful.

So, I built an AI food waste analyser using Microsoft Copilot’s agent feature. The setup work was the real effort here: I had to give the agent detailed context about every restaurant and galley on our ships – the dining format, typical covers per venue, and the operational differences between vessels in the fleet. I also had to be specific about what I wanted out the other end, including prompting it to produce a one-page synopsis suitable for senior leadership whenever monthly data is submitted.

The result is an agent that now takes our monthly food waste figures and breaks them down by venue, meal service (breakfast, lunch, dinner, afternoon tea, supper) and by source, distinguishing waste from plates, from the buffet, or from surplus in the galley. More usefully, it identifies trends: which venues are creeping upward, which meal services are improving, and where the data suggests a specific intervention would help. That’s the real value.

It took time to get here. Building the initial prompts and venue profiles was only the start, the harder part was checking, month after month, that the agent was interpreting the data correctly and that its trend calls matched what we’d expect from manual review. That validation period mattered. We’ve only started sharing the tool’s output beyond the environmental & sustainability team now that we’re confident in it, and that confidence came from repeated testing, not from the first output looking plausible.

Fuelling compliance
The second tool addresses a much newer challenge: FuelEU Maritime, the EU regulation that sets progressively tightening limits on the greenhouse gas intensity of energy used by ships. Having already worked through and understood the regulation’s requirements, the challenge was less about interpretation and more about turning those requirements into a practical decision-making tool that could support compliance planning across the fleet.

Using an AI coding tool, I built a FuelEU Maritime calculator through a process often described as “vibe coding” – defining in plain language the calculations, variables, and outputs required, and then refining the AI-generated code rather than writing it manually. The resulting calculator takes fleet fuel consumption data as input and calculates our current compliance position, the percentage of biofuel required to achieve compliance, and, most importantly for procurement planning, the exact metric tonnage of biofuel needed to meet the annual target.

Before adopting the tool, we validated its outputs against established manual calculations. The results matched, providing confidence that the calculations were being performed correctly. This verification step was essential; compliance tools must be demonstrably accurate before they can be relied upon for operational or commercial decisions. The calculator now provides a fast and reliable way to assess our FuelEU position.

What both tools have in common
Neither of these projects involved me writing code from scratch, and neither required a data science team. What they did require was domain expertise: an understanding of the finer details and context, translated carefully into instructions the AI could act on, followed by healthy scepticism about the output until it had been checked against something trustworthy. The food waste analyser only became useful once it could correctly separate plate waste from overproduction across dozens of different service contexts. The FuelEU calculator only became trustworthy once its numbers matched the manual method.

That’s the pattern I’d point to for anyone in maritime or another regulated, operationally complex industry considering similar tools: AI agents and AI-assisted coding lower the barrier to building something useful dramatically, but they don’t remove the need for subject-matter knowledge or for verification. The time saved doesn’t come from skipping the thinking, it comes from not having to build the underlying software/process yourself. In both cases here, the actual bottleneck was making sure the tool understood our operation well enough to be trusted, not writing the tool itself.
We’re now looking across the wider business to see how we can adopt AI solutions in a number of different departments. This same approach could apply, any process where the constraint has been “we understand the problem but don’t have the resource to build a tool for it” is worth a second look.

Words: Dominic Simpson, Environmental & Sustainability Lead at Fred Olsen Cruise Lines. Photo from Fred Olsen Cruise Lines.

——–
Like this article? Click here to sign up for our free monthly ‘De-Mystifying Digitalisation and AI in Maritime’ newsletter and get the latest cross-sector expert insights, news, interviews, videos and case studies straight to your inbox. You can also buy a ticket to the Demystifying AI in Maritime 2026 London Summit on our website here.